CN117557571A - Visual detection method and system for alloy resistance welding defects based on image enhancement - Google Patents

Visual detection method and system for alloy resistance welding defects based on image enhancement Download PDF

Info

Publication number
CN117557571A
CN117557571A CN202410045725.6A CN202410045725A CN117557571A CN 117557571 A CN117557571 A CN 117557571A CN 202410045725 A CN202410045725 A CN 202410045725A CN 117557571 A CN117557571 A CN 117557571A
Authority
CN
China
Prior art keywords
filtering
image
points
welding
point
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202410045725.6A
Other languages
Chinese (zh)
Other versions
CN117557571B (en
Inventor
涂振坤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Simei Microelectronics Technology Suzhou Co ltd
Original Assignee
Simei Microelectronics Technology Suzhou Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Simei Microelectronics Technology Suzhou Co ltd filed Critical Simei Microelectronics Technology Suzhou Co ltd
Priority to CN202410045725.6A priority Critical patent/CN117557571B/en
Publication of CN117557571A publication Critical patent/CN117557571A/en
Application granted granted Critical
Publication of CN117557571B publication Critical patent/CN117557571B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/10—Image enhancement or restoration using non-spatial domain filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/30—Noise filtering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20024—Filtering details
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30152—Solder
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Multimedia (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Quality & Reliability (AREA)
  • Image Processing (AREA)

Abstract

The invention relates to the technical field of image data processing, in particular to an alloy resistance welding defect visual detection method and system based on image enhancement, comprising the following steps: the method comprises the steps of obtaining a welding gray level image, screening out first filtering points, obtaining a plurality of neighborhood image blocks of each first filtering point, obtaining filtering weights of each neighborhood image block, performing first non-local mean value filtering processing to obtain a first filtering image, obtaining an enhanced image of the welding gray level image according to gray level values of pixel points in the first filtering image and filtering weights of all neighborhood image blocks of all first filtering points, and judging whether welding quality is qualified or not. According to the invention, the filtering weight of the neighborhood image block of the self-adaptive pixel point is used, and the filtering threshold value and the filtering weight in the filtering iteration process are continuously adjusted, so that the image filtering denoising effect is improved, the enhanced image with higher quality is obtained, and the accuracy of the welding defect detection result is improved.

Description

Visual detection method and system for alloy resistance welding defects based on image enhancement
Technical Field
The invention relates to the technical field of image data processing, in particular to an alloy resistance welding defect visual detection method and system based on image enhancement.
Background
Resistance welding is a welding method of joining alloy workpieces together by applying a current and voltage to the joint of the workpieces to generate heat, melting the alloy material and forming a weld. When the welding speed is too high or the current and voltage applied to the joint are too high, holes are generated in the welding seam area of the alloy workpiece, the existence of the holes can reduce the durability and the firmness of the alloy workpiece, and the image processing technology is currently used for detecting welding defects.
The existing problems are as follows: the acquisition of the welding image may be affected by serious noise interference to affect the defect detection result, and a non-local mean filtering algorithm is often used for noise filtering at present, but when the weight of a neighborhood image block and the selection of the filtering iteration times in the algorithm are inappropriate, the noise filtering effect of the image may be reduced, so that the image quality after noise filtering is still poorer, and the accuracy of the welding defect detection result based on image processing is reduced.
Disclosure of Invention
The invention provides an alloy resistance welding defect visual detection method and system based on image enhancement, which are used for solving the existing problems.
The invention discloses an alloy resistance welding defect visual detection method and system based on image enhancement, which adopts the following technical scheme:
One embodiment of the invention provides an image-enhancement-based visual detection method for alloy resistance welding defects, which comprises the following steps:
collecting an alloy resistance welding image, and carrying out graying treatment to obtain a welding gray image; in the welding gray level image, according to the gray level value of each pixel and the neighborhood pixel points, high-bright points and low-bright points are screened out from all the pixel points, and the filtering possibility of each pixel point is obtained; the pixel points with the filtering possibility larger than a preset screening threshold value are marked as first filtering points;
obtaining a plurality of neighborhood image blocks of each first filtering point by using a non-local mean filtering algorithm; obtaining the filtering weight of each neighborhood image block according to the gray value of the pixel points in each neighborhood image block, the distance between the pixel points and the number of high-bright points and low-bright points; according to the filtering weights of all neighborhood image blocks of all the first filtering points, carrying out first non-local mean filtering processing to obtain a first filtering image;
obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all neighborhood image blocks of all the first filtering points;
And judging whether the welding quality in the welding gray level image is qualified or not according to the enhanced image of the welding gray level image.
Further, in the welding gray level image, according to the gray level value of each pixel and the neighboring pixel, the high bright point and the low bright point are screened out from all the pixels, and the filtering possibility of each pixel is obtained, including the following specific steps:
the average value of gray values of all the pixels in the eight neighborhood of each pixel is recorded as the gray threshold value of each pixel in the welding gray image;
the pixel points with gray values larger than the gray threshold value in the welding gray image are marked as high-bright points;
the pixel point with the gray value smaller than or equal to the gray threshold value in the welding gray image is marked as a low bright point;
in the welding gray level image, calculating the distance between any two highlight points, and recording the average value of the distances between all highlight points as the noise discrete degree;
according to the noise discrete degree, the gray value and the gray threshold value of each pixel point in the welding gray image, a specific calculation formula corresponding to the filtering possibility of each pixel point in the welding gray image is obtained as follows:
；
wherein the method comprises the steps ofFor the filtering possibility of the x-th pixel point in the welding gray image,/for the welding gray image >For the gray value of the x-th pixel point in the welding gray image,/for the welding gray image>For the gray threshold value of the x-th pixel point in the welding gray image, A is the noise discrete degree, < ->Is a linear normalization function.
Further, the filtering weight of each neighborhood image block is obtained according to the gray value of the pixel points, the distance between the pixel points, the number of the high bright points and the low bright points in each neighborhood image block, and the specific steps are as follows:
recording any one first filtering point as a target point;
recording any one neighborhood image block of the target point as a target block;
in the target block, calculating the distance between each pixel point and each highlight point, and recording the average value of the distances between each pixel point and all the highlight points as the deviation distance of each pixel point;
dividing a target block into a plurality of connected domains by using a region growing algorithm;
according to the number of pixel points in all the connected domains in the target block, the deviation distance and gray value of all the pixel points, the number of high-bright points and low-bright points and the distance between the central point of the target block and the target point, a specific calculation formula corresponding to the filtering weight of the target block is obtained:
；
wherein Q is the filtering weight of the target block, Z is the noise level of the target block, B is the defect degree of the target block, r is the distance between the center point of the target block and the target point, m is the number of highlight points in the target block, Gray value for the i-th pixel in the target block,/>For the offset distance of the ith pixel point in the target block, n is the number of pixel points in the target block,/for the target block>For the number of low bright spots in the target block, +.>For the number of pixel points in the communication domain where the ith pixel point in the target block is located, +.>Is a linear normalization function.
Further, the first non-local mean filtering processing is performed according to the filtering weights of all the neighborhood image blocks of all the first filtering points to obtain a first filtered image, which comprises the following specific steps:
according to the filtering weights of all neighborhood image blocks of the target point, performing first non-local mean filtering processing by using a non-local mean filtering algorithm to obtain gray values of the target point after first filtering;
in the welding gray level image, the pixel point which is not the first filtering point is marked as a normal point;
and (3) recording an image formed by the gray values of all the first filtering points after the first filtering and the gray values of all the normal points as a first filtering image.
Further, the step of obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all the neighborhood image blocks of all the first filtering points comprises the following specific steps:
Obtaining a second screening threshold value of the first filtering image according to the times of the first filtering image subjected to non-local mean value filtering and the gray values of all pixel points;
screening out a plurality of second filtering points in the first filtering image according to the second screening threshold;
obtaining the filtering weight of each neighborhood image block of each second filtering point according to the filtering weight of the neighborhood image block of the first filtering point and the times of non-local mean value filtering processing of the first filtering image;
according to the filtering weights of all neighborhood image blocks of each second filtering point, performing second non-local mean value filtering processing to obtain a gray value of each second filtering point after second filtering;
in the first filtering image, the image formed by the gray values of all second filtering points after the second filtering and the gray values of all other pixels which are not the second filtering points is recorded as a second filtering image;
screening out a plurality of third filtering points in the second filtering image according to the acquisition mode of the second filtering points;
obtaining a third-time filtered gray value of each third-time filtering point according to the second-time filtered gray value obtaining mode of each second-time filtering point;
In the second time filtering image, the image formed by the gray values of all third time filtering points after third time filtering and the gray values of all other pixel points which are not the third time filtering points is recorded as a third time filtering image;
and similarly, stopping the non-local mean value filtering processing until the first random filtering point does not exist in the first random filtering image, and recording the first random filtering image as an enhanced image of the welding gray level image.
Further, the specific calculation formula corresponding to the second filtering threshold value of the first filtered image is obtained according to the number of times that the first filtered image is subjected to non-local mean value filtering processing and gray values of all pixel points, and is as follows:
；
wherein the method comprises the steps ofA second screening threshold value for the first filtered image,/->For a preset screening threshold, b is a preset constant, < ->For the number of times the first filtered image is subjected to non-local mean filtering +.>For the noise degree of the x-th pixel point in the first filtered image, y is the number of all pixel points in the first filtered image,/for the x-th pixel point>For the gray value of the x-th pixel point in the first filtered image, < >>For the average value of gray values of all pixel points in the first filtering image, < > >Is a linear normalization function.
Further, the step of screening out a plurality of second filtering points in the first filtering image according to the second filtering threshold includes the following specific steps:
according to the one-to-one correspondence between the first filtering image and the pixel points in the welding gray level image, the pixel points of the first filtering point corresponding to the first filtering image are marked as second filtering points to be filtered in the first filtering image;
obtaining the filtering possibility of each second filtering point in the first filtering image according to the obtaining mode of the filtering possibility of each pixel point;
and among all the second-time to-be-filtered points, the second-time to-be-filtered points with the filtering possibility larger than the second screening threshold value are marked as second-time filtering points.
Further, the filtering weight of each neighborhood image block of each second filtering point is obtained according to the filtering weight of the neighborhood image block of the first filtering point and the number of times that the first filtering image is subjected to non-local mean filtering processing, and the specific steps include:
obtaining a plurality of neighborhood image blocks of each second filtering point according to the acquisition mode of the plurality of neighborhood image blocks of each first filtering point;
Recording any second filtering point as a second target point;
recording any neighborhood image block of the second target point as a second target block;
according to the position of the second target block in the first filtering image, marking the neighborhood image block of the second target block corresponding to the same position of the welding gray level image as a reference block; the reference block is a neighborhood image block of a first filtering point corresponding to the second target block;
determining a filter weight adjustment value of the second target block according to the number of second filter points in the second target block, the times of non-local mean value filter processing of the first filter image and the filter weight of the reference block;
and subtracting the filter weight adjustment value of the second target block from the filter weight of the reference block, and recording the filter weight of the second target block.
Further, the step of judging whether the welding quality in the welding gray level image is qualified according to the enhanced image of the welding gray level image comprises the following specific steps:
obtaining a welding defect area in an enhanced image of the welding gray level image by using a deep neural network;
when a welding defect area exists in the enhanced image of the welding gray level image, judging that the welding quality in the welding gray level image is unqualified;
And when the enhanced image of the welding gray level image does not have a welding defect area, judging that the welding quality in the welding gray level image is qualified.
The invention also provides an image-enhancement-based visual detection system for alloy resistance welding defects, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program stored in the memory so as to realize the steps of the image-enhancement-based visual detection method for alloy resistance welding defects.
The technical scheme of the invention has the beneficial effects that:
in the embodiment of the invention, a welding gray level image is acquired, and first filtering points are screened out from all pixel points. The method improves the speed of filtering operation by primarily screening out the pixel points needing filtering. And obtaining a plurality of neighborhood image blocks of each first filtering point so as to obtain the filtering weight of each neighborhood image block, and carrying out first non-local mean value filtering processing according to the filtering weights of all neighborhood image blocks of all first filtering points so as to obtain a first filtering image. The filtering denoising effect is improved by reducing the filtering weight of the neighborhood image block with larger noise interference. And obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all the neighborhood image blocks of all the first filtering points, and further improving the filtering denoising effect by continuously adjusting the filtering threshold value and the filtering weights of the neighborhood image blocks in the filtering iterative process. And finally judging whether the welding quality in the welding gray level image is qualified or not. The filtering weight of the neighborhood image block of the self-adaptive pixel point is used, the filtering threshold value and the filtering weight in the filtering iteration process are continuously adjusted, the image filtering denoising effect is improved, the enhanced image with higher quality is obtained, and therefore the accuracy of the welding defect detection result is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of the steps of the visual inspection method of alloy resistance welding defects based on image enhancement of the invention;
FIG. 2 is a schematic diagram of a gray scale image of resistance welding of an alloy to be detected according to the present embodiment;
FIG. 3 is a schematic view of a first filtered image of a welding gray scale image according to the present embodiment;
FIG. 4 is a schematic view of an enhanced image of a welding gray scale image according to the present embodiment;
fig. 5 is a schematic diagram of a conventional filtered enhanced image of a welding gray scale image according to the present embodiment.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following detailed description refers to the specific implementation, structure, characteristics and effects of the method and the system for visual detection of alloy resistance welding defects based on image enhancement according to the invention, which are provided by the invention, with reference to the accompanying drawings and the preferred embodiment. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the image-enhancement-based alloy resistance welding defect visual detection method and system provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a visual inspection method for an alloy resistance welding defect based on image enhancement according to an embodiment of the invention is shown, the method includes the following steps:
step S001: collecting an alloy resistance welding image, and carrying out graying treatment to obtain a welding gray image; in the welding gray level image, according to the gray level value of each pixel and the neighborhood pixel points, high-bright points and low-bright points are screened out from all the pixel points, and the filtering possibility of each pixel point is obtained; and (5) marking the pixel points with the filtering possibility larger than a preset screening threshold value as first filtering points.
The main purpose of this embodiment is to obtain a high-quality alloy resistance welding image by an image processing technique, thereby accurately detecting whether or not there is a welding defect. Because the acquired image may be interfered by serious noise to influence the defect detection result, the non-local mean filtering algorithm is improved, the filtering denoising effect is improved, and a high-quality image is obtained.
What needs to be described is: the non-local mean filtering algorithm is a well-known technique, and the specific method is not described here. Non-local mean filtering typically performs multiple iterations to progressively reduce noise in the image and achieve better smoothing. The general process of the algorithm is as follows: selecting a search window for each pixel point in an image, obtaining a plurality of neighborhood image blocks corresponding to the pixel point in the search window, calculating the weight of each neighborhood image block, carrying out weighted average to obtain a gray value after filtering the pixel point, and repeating the filtering process until a set stopping criterion is met, thereby completing the image smoothing denoising process.
And collecting an alloy resistance welding image, and carrying out graying treatment to obtain a welding gray image. Fig. 2 is a schematic diagram of a gray scale image of resistance welding of an alloy to be detected according to the present embodiment.
What needs to be described is: in this embodiment, an industrial camera is used to collect an alloy resistance welding image in a top view, and the gray-scale processing of the image is a known technology, and a specific method is not described here.
When filtering an image, when more noise points exist around a certain pixel point in the image, it is difficult to filter the noise points in the image through one-time filtering, and a large number of surrounding noise points are often removed through multiple times of filtering processing. Therefore, when the difference between a certain pixel point and the surrounding pixel points is large, more filtering times are usually needed, so as to ensure the filtering effect. For the image, the more the filtering times, the fewer the number of pixels to be filtered, so as to obtain a better filtering effect, but the more the filtering times, the excessive smoothing of the image may be caused, and the detail information is lost, so that the embodiment improves the filtering denoising effect by adjusting the parameters in the filtering iterative process. The first non-local mean filtering process is as follows:
The pixel characteristics of the defect area and the noise area in the image are different from those of other areas, the gray values of the pixels in the image can be analyzed, so that the possibility that the pixels in the image belong to defects or noise points is calculated, the pixels with high possibility are set as first filtering points of the image, and filtering analysis is carried out on the pixels.
It is known that defects in a welded region in an image appear as regions of low gray scale, and therefore the lower the gray scale value of a pixel in the image, the higher the likelihood that it belongs to a defective region. Meanwhile, the defect on the welding line appears as a low gray scale region, so that if the gray scale value of the neighborhood of the low gray scale pixel point is lower, the probability that the defect belongs to the defect region is higher. The noise point in the current scene appears as a highlighted discrete pixel in the image, and if the higher the gray value of the pixel is and the farther the distance between the highlighted pixels is, the higher the probability that the noise point belongs to noise.
In the welding gray level image, the average value of gray level values of all pixels in eight adjacent domains of each pixel is recorded as the gray level threshold value of each pixel in the welding gray level image.
In the welding gray level image, the pixel point with the gray level value larger than the gray level threshold value of the pixel point is marked as a high bright point.
In the welding gray level image, the pixel point with the gray level value smaller than or equal to the gray level threshold value of the pixel point is marked as a low bright point.
And calculating the distance between any two highlight points, and recording the average value of the distances between all highlight points as the noise discrete degree.
Taking the x-th pixel in the welding gray image as an example, the calculation formula of the filtering possibility of the x-th pixel in the welding gray image is as follows:
；
wherein the method comprises the steps ofFor the filtering possibility of the x-th pixel point in the welding gray image,/for the welding gray image>For the gray value of the x-th pixel point in the welding gray image,/for the welding gray image>For the gray threshold value of the x-th pixel point in the welding gray image, A is the noise discrete degree, < ->Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval.
What needs to be described is:the larger the x pixel point is, the darker the local area is, the more likely the x pixel point is in a defect area, the more filtering and denoising processing is needed, and the more>The larger the x-th pixel point is, the more likely the x-th pixel point is a noise point, and the more filtering and denoising processing is needed, so the x-th pixel point is used>Representing the filtering probability of the x-th pixel.
The filtering possibility of each pixel point is obtained in the above manner.
The preset screening threshold value in this embodiment is 0.5, which is described as an example, and other values may be set in other embodiments, which is not limited in this embodiment.
And (5) marking the pixel points with the filtering possibility larger than a preset screening threshold value as first filtering points.
Step S002: obtaining a plurality of neighborhood image blocks of each first filtering point by using a non-local mean filtering algorithm; obtaining the filtering weight of each neighborhood image block according to the gray value of the pixel points in each neighborhood image block, the distance between the pixel points and the number of high-bright points and low-bright points; and carrying out first non-local mean filtering processing according to the filtering weights of all neighborhood image blocks of all the first filtering points to obtain a first filtering image.
When a larger noise level exists in the neighborhood image block, the lower the weight of the neighborhood image block during filtering is, so that the influence of noise on the filtering process is reduced. When the pixel points in the neighborhood image block have higher expression degree on the defect area, the neighborhood image block has higher weight when filtering, so that the defect area can be better identified. And determining the filtering weights corresponding to the different neighborhood image blocks to weight according to the pixel point performance in the neighborhood image blocks and combining the distance between the neighborhood image blocks and the first filtering point to obtain a first filtering result.
Any one of the first filtering points is recorded as a target point.
The preset search window side length of this embodiment is 21, and the preset neighborhood image block side length is 5, which is described as an example, but other values may be set in other embodiments, which is not limited in this embodiment.
And obtaining a plurality of neighborhood image blocks of the target point by using a non-local mean filtering algorithm according to the preset search window side length and the preset neighborhood image block side length.
What needs to be described is: the "neighborhood image block" of the pixel point in the non-local mean filtering algorithm is also called "neighborhood block", "pixel block", etc., and the specific acquisition process of the plurality of neighborhood image blocks of the target point in this embodiment is: in the welding gray level image, the target point is taken as the center to constructIn which a number of +.>Is known in the art.
And marking any neighborhood image block of the target point as a target block.
It is known that the noise level in a target block is higher when the noise performance of pixels in the target block is greater and the number of pixels that may belong to noise is greater. The higher the noise level in the target block, the greater the degree of influence of the gray values of the pixels in the target block on the filtering result when filtering is performed, and therefore the lower the corresponding filtering weight when the noise level in the target block is higher.
When the target block exhibits a higher degree of defective area, the target block obtains a larger weight when filtering is performed, so that the first filtering point is closer to the defective area. And analyzing the low-bright spots in the target block, and when the gray value of any one pixel point is lower and the area of the local area with similar gray value is larger, the probability of the local area belonging to the defect is higher, namely the defect area is more represented. Traversing all the pixel points, and calculating the defect degree of the target block according to the number of the low-light points in the target block and the representing degree of the pixel points in the target block to the defect region.
And in the target block, calculating the distance between each pixel point and each highlight point, and recording the average value of the distances between each pixel point and all the highlight points as the deviation distance of each pixel point.
What needs to be described is: if there is no highlight point in the target block, the offset distance of each pixel point in the target block is 0.
In the target block, the target block is divided into a plurality of connected domains by using a region growing algorithm. The region growing algorithm is a well-known technique, and a specific method is not described herein.
The calculation formula of the filtering weight Q of the target block can be known as follows:
；
Wherein Q is the filtering weight of the target block, Z is the noise level of the target block, B is the defect degree of the target block, r is the distance between the center point of the target block and the target point, m is the number of highlight points in the target block,gray value for the i-th pixel in the target block,/>For the offset distance of the ith pixel point in the target block, n is the number of pixel points in the target block,/for the target block>For the number of low bright spots in the target block, +.>For the number of pixel points in the communication domain where the ith pixel point in the target block is located, +.>Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval.
What needs to be described is: according to the analysis, the target block should be close to the defect area, so when the number m of highlight points in the target block is larger and the gray value of the pixel point is larger, the noise influence existing in the target block is larger, whenThe larger the distribution of the highlight pixel points is, the more discrete the distribution of the highlight pixel points is, the more the possibility of belonging to noise is, therefore +.>The greater Z, the smaller the filter weight of the target block, which represents the noise level of the target block. The lower the gray value of a pixel, the higher the likelihood that the pixel belongs to a defective area,the larger the area of the connected region where the pixel point is located, the higher the probability that the region belongs to the defect region, i.e. +. >The larger the target block, the greater the possibility of defects in the target block, and +.>The larger indicates the more defective pixels in the target block, thus using +.>The greater B, the more important the target block, the greater the filtering weight is required. When r is smaller, the closer the target block is to the target point, the more reliable the target block is, the greater the filtering weight should be, thus using +.>Representing the filtering weights of the target block.
And obtaining the filtering weight of each neighborhood image block of the target point according to the mode.
And carrying out first non-local mean value filtering processing by using a non-local mean value filtering algorithm according to the filtering weights of all the neighborhood image blocks of the target point to obtain a gray value of the target point after the first filtering.
According to the mode, the gray value after the first filtering of each first filtering point is obtained.
In the welding gray scale image, a pixel point which is not the first filtering point is denoted as a normal point.
The image formed by the gray values after the first filtering of all the first filtering points and the gray values of all the normal points is recorded as the first filtering image, and fig. 3 is a schematic diagram of the first filtering image of the welding gray image provided by the embodiment. Thus, the first non-local mean value filtering processing of the welding gray level image is completed.
Step S003: and obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all neighborhood image blocks of all the first filtering points.
The number of pixels in the image, which meets the condition of the filtering point, gradually decreases with increasing filtering times, so that the condition of selecting the point to be filtered with increasing filtering times should also be gradually strict. For the filtered point filtered again, the weight of the neighborhood image block filtered last time cannot be used as the current weight to be filtered, and self-adaption needs to be carried out again. The second non-local mean filtering process is as follows:
and filtering the possible filtering points in the image based on the filtering result of the first time according to the filtering conditions of the filtering points, wherein in order to make the pixel representation of the defect area in the image more obvious and reduce the noise in the image, the filtering conditions of the filtering points need to be self-adaptive according to the filtering times, namely, when the filtering times are increased, the filtering conditions of the filtering points should be gradually increased, and when the noise level in the image is higher, the filtering conditions of the filtering points are also larger.
The second screening threshold of the first filtered image is knownThe calculation formula of (2) is as follows:
；
wherein the method comprises the steps ofA second screening threshold value for the first filtered image,/->For a preset screening threshold, b is a preset constant, < ->For the number of times the first filtered image is subjected to non-local mean filtering, it is to be noted that currently +.>1, during the third non-local mean filtering, +.>The number of times the second filtered image is subjected to non-local mean filtering, i.e. +.>2./>For the noise degree of the x-th pixel point in the first filtered image, y is the number of all pixel points in the first filtered image,/for the x-th pixel point>For the gray value of the x-th pixel point in the first filtered image, < >>For the average value of gray values of all pixel points in the first filtering image, < >>Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval. In this embodiment, b is set to 0.5, which is described as an example, and other values may be set in other embodiments, and the embodiment is not limited thereto.
What needs to be described is: it is known that, with iteration of non-local mean filtering, the filtering point screening threshold should be more and more stringent,representing the degree of dispersion of the xth pixel point in the first filtering image in the image, wherein the larger the degree of dispersion is, the more the possibility of noise is, so +. >Representing the noise level of the first filtered image, therefore +.>The larger the more the screening threshold needs to be raised, therefore +.>Representing a second screening threshold for the first filtered image.
And according to the one-to-one correspondence between the first filtering image and the pixel points in the welding gray level image, the pixel points corresponding to the first filtering points in the first filtering image are marked as second filtering points.
And obtaining the filtering possibility of each second point to be filtered in the first filtering image according to the obtaining mode of the filtering possibility of each pixel point in the welding gray level image.
And marking the second time to-be-filtered point with the filtering possibility larger than the second time screening threshold value as a second time filtering point.
Then the filter weights of the neighborhood image blocks of the second filtering point need to be obtained. As the number of filtering increases, the number of pixels meeting the filtering condition in the image should decrease gradually, so that when the number of filtering in the neighborhood image block increases, the filtering effect of the neighborhood image block is still worse after multiple filtering, and the weight of the neighborhood image block as a filtering reference should decrease.
And obtaining a plurality of neighborhood image blocks of each second filtering point according to the acquisition modes of the plurality of neighborhood image blocks of each first filtering point.
Any one of the second filtering points is recorded as a second target point.
And marking any neighborhood image block of the second target point as a second target block.
And according to the one-to-one correspondence between the first filtering image and the pixel points in the welding gray level image, the neighborhood image block of the second target block in the welding gray level image is marked as a reference block. The reference block is a neighborhood image block of the first filtering point corresponding to the second target block.
From this, the filtering weights of the second order target block can be knownThe calculation formula of (2) is as follows:
；
wherein the method comprises the steps ofFiltering weights for the second order target block, < ->Filter weights for reference block, +.>For the number of times the first filtered image is subjected to non-local mean filtering +.>For the number of pixels in the second sub-target block, is->Is the number of second order filter points in the second order target block. />Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval.
What needs to be described is: the second filtering point is selected from the first filtering point, so that the reference block corresponding to the second target block must obtain the filtering weight of the reference block in the first non-local mean filtering process. When (when)And->When the target block is larger, the filtering effect of the second target block is still poor after the second target block is filtered for multiple times, so that noise still exists in the second target block, and the filtering weight of the second target block is reduced, so that the second target block is used +. >Representing the filter weight adjustment value of the second order target block. Wherein (1)>Representation->Is thus +.>Representing the filtering weights of the second order target block.
And obtaining the filtering weight of each neighborhood image block of the second target point in the mode.
And carrying out second non-local mean value filtering processing by using a non-local mean value filtering algorithm according to the filtering weights of all neighborhood image blocks of the second target point to obtain a gray value of the second target point after the second filtering.
And obtaining the gray value after the second filtering of each second filtering point according to the mode.
In the first-time filtered image, an image formed by gray values of all second-time filtered points and gray values of all other pixels which are not the second-time filtered points is recorded as a second-time filtered image.
The third non-local mean filtering process is known as follows:
obtaining a third screening threshold value of the second filtering image according to the obtaining mode of the second screening threshold value of the first filtering image。
What needs to be described is:according to the difference of gray values of pixel points in the second filtered image, the method is to +.>And (5) performing adjustment.
And according to the one-to-one correspondence between the pixel points in the first filtering image and the second filtering image, the pixel points of the second filtering point corresponding to the second filtering image are marked as third filtering points.
And obtaining the filtering possibility of each third filtering point in the second filtering image according to the obtaining mode of the filtering possibility of each pixel point in the welding gray level image.
And marking the third time to-be-filtered point with the filtering possibility larger than the third time screening threshold value as the third time filtering point.
And obtaining the neighborhood image block of the third filtering point according to the acquisition mode of the neighborhood image block of the first filtering point.
And obtaining the filter weight of each neighborhood image block of each third filtering point according to the filter weight obtaining mode of each neighborhood image block of each second filtering point.
What needs to be described is: the filter weights of the neighborhood image blocks of the third filtering point are obtained on the filter weights of the neighborhood image blocks of the second filtering point.
And carrying out third non-local mean value filtering processing by using a non-local mean value filtering algorithm according to the filtering weights of all neighborhood image blocks of each third filtering point to obtain a gray value after third filtering of each third filtering point.
In the second time filtered image, the image formed by the gray values of all third time filtered points and the gray values of all other pixels which are not the third time filtered points is recorded as a third time filtered image.
Similarly, until the non-local mean value filtering process is stopped when the arbitrary filtering point does not exist in the arbitrary filtering image, the arbitrary filtering image is recorded as an enhanced image of the welding gray image, and fig. 4 is a schematic diagram of the enhanced image of the welding gray image provided in the embodiment. The existing filtering algorithm is used for filtering the alloy resistance welding gray level image to be detected to obtain an existing filtering enhancement image of the welding gray level image, and fig. 5 is a schematic diagram of the existing filtering enhancement image of the welding gray level image provided by the embodiment. The filtering image processed by the method, namely the enhancement image of the welding gray level image, is compared with the existing filtering enhancement image of the welding gray level image, namely the filtering image obtained by the existing algorithm, so that the filtering result of the method can be found to have the characteristics of clearer and better filtering effect, and the visual detection of the welding defect region of the involution Jin Dianzu can be realized more conveniently.
What needs to be described is: in this embodiment, the non-local mean filtering is performed 10 times at most, and this is described as an example, but other values may be set in other embodiments, and this embodiment is not limited thereto.
Step S004: and judging whether the welding quality in the welding gray level image is qualified or not according to the enhanced image of the welding gray level image.
The embodiment of the invention adopts a deep neural network to identify the welding defect area in the enhanced image of the split welding gray level image.
The relevant content of the deep neural network is as follows:
the deep neural network used in this embodiment is a deep labv3 neural network; the dataset used is an enhanced image dataset of the welding gray scale image.
The pixel points to be segmented are divided into 2 classes, namely, the labeling process of the corresponding label of the training set is as follows: and the single-channel semantic label is marked as 0 corresponding to the pixel points in the positions belonging to the background class, and is marked as 1 corresponding to the welding defect area. This process is well known, and detailed description is omitted in this embodiment.
The task of the network is classification, so the loss function used is a cross entropy loss function.
And obtaining a welding defect area in the enhanced image of the welding gray level image through the deep neural network.
And when a welding defect area exists in the enhanced image of the welding gray level image, judging that the welding quality in the welding gray level image is unqualified.
And when the enhanced image of the welding gray level image does not have a welding defect area, judging that the welding quality in the welding gray level image is qualified.
The present invention has been completed.
In summary, in the embodiment of the present invention, the welding gray level image is obtained, and the first filtering point is selected from all the pixel points. And acquiring a plurality of neighborhood image blocks of each first filtering point, thereby obtaining the filtering weight of each neighborhood image block. And carrying out first non-local mean filtering processing according to the filtering weights of all neighborhood image blocks of all the first filtering points to obtain a first filtering image. And obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all neighborhood image blocks of all the first filtering points, so as to judge whether the welding quality in the welding gray level image is qualified. According to the invention, the filtering weight of the neighborhood image block of the self-adaptive pixel point is used, and the filtering threshold value and the filtering weight in the filtering iteration process are continuously adjusted, so that the image filtering denoising effect is improved, the enhanced image with higher quality is obtained, and the accuracy of the welding defect detection result is improved.
The invention also provides an image-enhancement-based visual detection system for alloy resistance welding defects, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program stored in the memory so as to realize the steps of the image-enhancement-based visual detection method for alloy resistance welding defects.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the invention, but any modifications, equivalent substitutions, improvements, etc. within the principles of the present invention should be included in the scope of the present invention.

Claims (10)

1. The visual detection method for the alloy resistance welding defect based on image enhancement is characterized by comprising the following steps of:
collecting an alloy resistance welding image, and carrying out graying treatment to obtain a welding gray image; in the welding gray level image, according to the gray level value of each pixel and the neighborhood pixel points, high-bright points and low-bright points are screened out from all the pixel points, and the filtering possibility of each pixel point is obtained; the pixel points with the filtering possibility larger than a preset screening threshold value are marked as first filtering points;
obtaining a plurality of neighborhood image blocks of each first filtering point by using a non-local mean filtering algorithm; obtaining the filtering weight of each neighborhood image block according to the gray value of the pixel points in each neighborhood image block, the distance between the pixel points and the number of high-bright points and low-bright points; according to the filtering weights of all neighborhood image blocks of all the first filtering points, carrying out first non-local mean filtering processing to obtain a first filtering image;
Obtaining an enhanced image of the welding gray level image according to the gray level value of the pixel point in the first filtering image and the filtering weights of all neighborhood image blocks of all the first filtering points;
and judging whether the welding quality in the welding gray level image is qualified or not according to the enhanced image of the welding gray level image.
2. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 1, wherein in the welding gray level image, according to gray level values of each pixel and its neighborhood pixel points, high bright points and low bright points are screened out from all the pixel points, and filtering possibility of each pixel point is obtained, comprising the following specific steps:
the average value of gray values of all the pixels in the eight neighborhood of each pixel is recorded as the gray threshold value of each pixel in the welding gray image;
the pixel points with gray values larger than the gray threshold value in the welding gray image are marked as high-bright points;
the pixel point with the gray value smaller than or equal to the gray threshold value in the welding gray image is marked as a low bright point;
in the welding gray level image, calculating the distance between any two highlight points, and recording the average value of the distances between all highlight points as the noise discrete degree;
According to the noise discrete degree, the gray value and the gray threshold value of each pixel point in the welding gray image, a specific calculation formula corresponding to the filtering possibility of each pixel point in the welding gray image is obtained as follows:
；
wherein the method comprises the steps ofFor the filtering possibility of the x-th pixel point in the welding gray image,/for the welding gray image>For the gray value of the x-th pixel point in the welding gray image,/for the welding gray image>For the gray threshold value of the x-th pixel point in the welding gray image, A is the noise discrete degree, < ->Is a linear normalization function.
3. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 1, wherein the obtaining the filtering weight of each neighborhood image block according to the gray value of the pixel points, the distance between the pixel points, the number of high bright points and low bright points in each neighborhood image block comprises the following specific steps:
recording any one first filtering point as a target point;
recording any one neighborhood image block of the target point as a target block;
in the target block, calculating the distance between each pixel point and each highlight point, and recording the average value of the distances between each pixel point and all the highlight points as the deviation distance of each pixel point;
Dividing a target block into a plurality of connected domains by using a region growing algorithm;
according to the number of pixel points in all the connected domains in the target block, the deviation distance and gray value of all the pixel points, the number of high-bright points and low-bright points and the distance between the central point of the target block and the target point, a specific calculation formula corresponding to the filtering weight of the target block is obtained:
；
wherein Q is the filtering weight of the target block, Z is the noise level of the target block, B is the defect degree of the target block, r is the distance between the center point of the target block and the target point, m is the number of highlight points in the target block,gray value for the i-th pixel in the target block,/>For the offset distance of the ith pixel point in the target block, n is the number of pixel points in the target block,/for the target block>For the number of low bright spots in the target block, +.>For the number of pixel points in the communication domain where the ith pixel point in the target block is located, +.>Is a linear normalization function.
4. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 3, wherein the first non-local mean filtering process is performed according to the filtering weights of all the neighborhood image blocks of all the first filtering points to obtain a first filtered image, comprising the following specific steps:
According to the filtering weights of all neighborhood image blocks of the target point, performing first non-local mean filtering processing by using a non-local mean filtering algorithm to obtain gray values of the target point after first filtering;
in the welding gray level image, the pixel point which is not the first filtering point is marked as a normal point;
and (3) recording an image formed by the gray values of all the first filtering points after the first filtering and the gray values of all the normal points as a first filtering image.
5. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 1, wherein the obtaining the enhancement image of the welding gray level image according to the gray level values of the pixel points in the first filtering image and the filtering weights of all the neighborhood image blocks of all the first filtering points comprises the following specific steps:
obtaining a second screening threshold value of the first filtering image according to the times of the first filtering image subjected to non-local mean value filtering and the gray values of all pixel points;
screening out a plurality of second filtering points in the first filtering image according to the second screening threshold;
obtaining the filtering weight of each neighborhood image block of each second filtering point according to the filtering weight of the neighborhood image block of the first filtering point and the times of non-local mean value filtering processing of the first filtering image;
According to the filtering weights of all neighborhood image blocks of each second filtering point, performing second non-local mean value filtering processing to obtain a gray value of each second filtering point after second filtering;
in the first filtering image, the image formed by the gray values of all second filtering points after the second filtering and the gray values of all other pixels which are not the second filtering points is recorded as a second filtering image;
screening out a plurality of third filtering points in the second filtering image according to the acquisition mode of the second filtering points;
obtaining a third-time filtered gray value of each third-time filtering point according to the second-time filtered gray value obtaining mode of each second-time filtering point;
in the second time filtering image, the image formed by the gray values of all third time filtering points after third time filtering and the gray values of all other pixel points which are not the third time filtering points is recorded as a third time filtering image;
and similarly, stopping the non-local mean value filtering processing until the first random filtering point does not exist in the first random filtering image, and recording the first random filtering image as an enhanced image of the welding gray level image.
6. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 5, wherein the specific calculation formula corresponding to the second filtering threshold value of the first filtered image is obtained according to the number of times of non-local mean value filtering processing of the first filtered image and gray values of all pixel points, and is as follows:
；
wherein the method comprises the steps ofA second screening threshold value for the first filtered image,/->For a preset screening threshold, b is a preset constant, < ->For the number of times the first filtered image is subjected to non-local mean filtering +.>For the noise degree of the x-th pixel point in the first filtered image, y is the number of all pixel points in the first filtered image,/for the x-th pixel point>For the gray value of the x-th pixel point in the first filtered image, < >>For all of the first filtered imagesMean value of gray values of pixel points, +.>Is a linear normalization function.
7. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 5, wherein the step of screening out a plurality of second filtering points in the first filtered image according to the second screening threshold comprises the following specific steps:
according to the one-to-one correspondence between the first filtering image and the pixel points in the welding gray level image, the pixel points of the first filtering point corresponding to the first filtering image are marked as second filtering points to be filtered in the first filtering image;
Obtaining the filtering possibility of each second filtering point in the first filtering image according to the obtaining mode of the filtering possibility of each pixel point;
and among all the second-time to-be-filtered points, the second-time to-be-filtered points with the filtering possibility larger than the second screening threshold value are marked as second-time filtering points.
8. The visual inspection method of alloy resistance welding defects based on image enhancement according to claim 5, wherein the obtaining the filtering weight of each neighborhood image block of each second filtering point according to the filtering weight of the neighborhood image block of the first filtering point and the number of times that the first filtering image is subjected to non-local mean value filtering processing comprises the following specific steps:
obtaining a plurality of neighborhood image blocks of each second filtering point according to the acquisition mode of the plurality of neighborhood image blocks of each first filtering point;
recording any second filtering point as a second target point;
recording any neighborhood image block of the second target point as a second target block;
according to the position of the second target block in the first filtering image, marking the neighborhood image block of the second target block corresponding to the same position of the welding gray level image as a reference block; the reference block is a neighborhood image block of a first filtering point corresponding to the second target block;
Determining a filter weight adjustment value of the second target block according to the number of second filter points in the second target block, the times of non-local mean value filter processing of the first filter image and the filter weight of the reference block;
and subtracting the filter weight adjustment value of the second target block from the filter weight of the reference block, and recording the filter weight of the second target block.
9. The visual inspection method for alloy resistance welding defects based on image enhancement according to claim 1, wherein the step of judging whether the welding quality in the welding gray level image is acceptable or not according to the enhancement image of the welding gray level image comprises the following specific steps:
obtaining a welding defect area in an enhanced image of the welding gray level image by using a deep neural network;
when a welding defect area exists in the enhanced image of the welding gray level image, judging that the welding quality in the welding gray level image is unqualified;
and when the enhanced image of the welding gray level image does not have a welding defect area, judging that the welding quality in the welding gray level image is qualified.
10. An image-enhanced visual inspection system for alloy resistance welding defects, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program when executed by the processor implements the steps of the image-enhanced visual inspection method for alloy resistance welding defects as claimed in any one of claims 1 to 9.
CN202410045725.6A 2024-01-12 2024-01-12 Visual detection method and system for alloy resistance welding defects based on image enhancement Active CN117557571B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410045725.6A CN117557571B (en) 2024-01-12 2024-01-12 Visual detection method and system for alloy resistance welding defects based on image enhancement

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410045725.6A CN117557571B (en) 2024-01-12 2024-01-12 Visual detection method and system for alloy resistance welding defects based on image enhancement

Publications (2)

Publication Number Publication Date
CN117557571A true CN117557571A (en) 2024-02-13
CN117557571B CN117557571B (en) 2024-03-26

Family

ID=89817080

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410045725.6A Active CN117557571B (en) 2024-01-12 2024-01-12 Visual detection method and system for alloy resistance welding defects based on image enhancement

Country Status (1)

Country Link
CN (1) CN117557571B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117876379A (en) * 2024-03-13 2024-04-12 山东鑫国矿业技术开发有限公司 Intelligent anchor rod defect detection method based on image characteristics
CN117893540A (en) * 2024-03-18 2024-04-16 乳山市创新新能源科技有限公司 Roundness intelligent detection method and system for pressure container
CN118014886A (en) * 2024-04-10 2024-05-10 中建安装集团有限公司 Ultrasonic pipeline crack image enhancement method
CN118052730A (en) * 2024-02-27 2024-05-17 青岛艺高林通林木研发有限公司 Northern colored leaf shrub seed screening method
CN118133053A (en) * 2024-05-08 2024-06-04 山东瑞福锂业有限公司 Industrial data collecting and processing system and method
CN118154599A (en) * 2024-05-10 2024-06-07 安达斯科技（大连）有限公司 Rubber and metal material bonding surface treatment quality identification method
CN118172738A (en) * 2024-05-16 2024-06-11 山东力加力钢结构有限公司 Traffic engineering road condition detection method based on machine vision

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116188462A (en) * 2023-04-24 2023-05-30 深圳市翠绿贵金属材料科技有限公司 Noble metal quality detection method and system based on visual identification
CN116542982A (en) * 2023-07-07 2023-08-04 山东中泳电子股份有限公司 Departure judgment device defect detection method and device based on machine vision

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116188462A (en) * 2023-04-24 2023-05-30 深圳市翠绿贵金属材料科技有限公司 Noble metal quality detection method and system based on visual identification
CN116542982A (en) * 2023-07-07 2023-08-04 山东中泳电子股份有限公司 Departure judgment device defect detection method and device based on machine vision

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118052730A (en) * 2024-02-27 2024-05-17 青岛艺高林通林木研发有限公司 Northern colored leaf shrub seed screening method
CN118052730B (en) * 2024-02-27 2024-07-30 青岛艺高林通林木研发有限公司 Northern colored leaf shrub seed screening method
CN117876379A (en) * 2024-03-13 2024-04-12 山东鑫国矿业技术开发有限公司 Intelligent anchor rod defect detection method based on image characteristics
CN117876379B (en) * 2024-03-13 2024-05-24 山东鑫国矿业技术开发有限公司 Intelligent anchor rod defect detection method based on image characteristics
CN117893540A (en) * 2024-03-18 2024-04-16 乳山市创新新能源科技有限公司 Roundness intelligent detection method and system for pressure container
CN117893540B (en) * 2024-03-18 2024-05-31 乳山市创新新能源科技有限公司 Roundness intelligent detection method and system for pressure container
CN118014886A (en) * 2024-04-10 2024-05-10 中建安装集团有限公司 Ultrasonic pipeline crack image enhancement method
CN118133053A (en) * 2024-05-08 2024-06-04 山东瑞福锂业有限公司 Industrial data collecting and processing system and method
CN118154599A (en) * 2024-05-10 2024-06-07 安达斯科技（大连）有限公司 Rubber and metal material bonding surface treatment quality identification method
CN118154599B (en) * 2024-05-10 2024-07-09 安达斯科技（大连）有限公司 Rubber and metal material bonding surface treatment quality identification method
CN118172738A (en) * 2024-05-16 2024-06-11 山东力加力钢结构有限公司 Traffic engineering road condition detection method based on machine vision

Also Published As

Publication number Publication date
CN117557571B (en) 2024-03-26

Similar Documents

Publication Publication Date Title
CN117557571B (en) Visual detection method and system for alloy resistance welding defects based on image enhancement
CN109242853B (en) PCB defect intelligent detection method based on image processing
CN115330758B (en) Welding quality detection method based on denoising treatment
CN115330645B (en) Welding image enhancement method
CN111612817A (en) Target tracking method based on depth feature adaptive fusion and context information
CN111815555A (en) Metal additive manufacturing image detection method and device combining anti-neural network with local binary
CN109472788B (en) Method for detecting flaw on surface of airplane rivet
CN116485797B (en) Artificial intelligence-based paint color difference rapid detection method
CN115311484B (en) Optimized clustering segmentation method for weld data
CN116363126B (en) Welding quality detection method for data line USB plug
CN117058147B (en) Environment-friendly plastic product defect detection method based on computer vision
CN109255799B (en) Target tracking method and system based on spatial adaptive correlation filter
CN117994154B (en) Intelligent image denoising method based on sensor
CN115841434A (en) Infrared image enhancement method for gas concentration analysis
CN115471503B (en) Equipment abnormity detection method for numerical control ingot splitting machine
CN117576086B (en) Method and system for detecting metal electron beam welding defects
CN117541582B (en) IGBT insulation quality detection method for high-frequency converter
CN117689574B (en) Medical image processing method for tumor radio frequency ablation diagnosis and treatment positioning
CN114862802A (en) Steel structure cutting control method based on image processing
CN118037715B (en) Intelligent welding spot analysis optimization method and system for PCBA board
CN111709964A (en) PCBA target edge detection method
CN111667509B (en) Automatic tracking method and system for moving target under condition that target and background colors are similar
CN115578390B (en) Welding control method for deaerator
CN113705672A (en) Threshold value selection method, system and device for image target detection and storage medium
CN118297839B (en) Transformer winding state detection method based on vision

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant